hi Quang -- I recommend clearing out your CMake temporary files after
making any conda environment changes. If you activate a different
conda environment, CMake will not know to recompute variables related
to Python's header files and libraries. So it might have been that you
invoked CMake with Python 2 activated and later activated Python 3
- Wes
On Tue, May 15, 2018 at 5:15 AM, Quang Vu <vhqu...@gmail.com> wrote:
> Yes Antoine, that happens when compiling Arrow under an activated conda
> environment.
> Thank you for all the info you are helping me with!
>
> Quang.
>
> On Mon, May 14, 2018 at 3:34 PM Antoine Pitrou <anto...@python.org> wrote:
>
>>
>> To give a bit more insight: you should compile Arrow with your conda
>> environment activated, so that it picks the right Python version (3.6.5,
>> in your case). If it's still picking the wrong Python version, that
>> might be a bug.
>>
>> Regards
>>
>> Antoine.
>>
>>
>> Le 14/05/2018 à 20:50, Quang Vu a écrit :
>> > Thanks Antoine,
>> >
>> > I will need to learn more about the compiling process that happens on my
>> > Mac, to see how that link to Python 2.
>> > I am not familiar with that process. But this is a good pointer for my
>> > issue. Thank you for your response to my issue!
>> >
>> > Quang.
>> >
>> > On Mon, May 14, 2018 at 12:50 PM Antoine Pitrou <anto...@python.org>
>> wrote:
>> >
>> >>
>> >> Hi Quang,
>> >>
>> >> It sounds like you have compiled Arrow against a Python 2 install but
>> >> are now trying to use it with Python 3. This won't work, the same
>> >> Python version must be used when compiling and when using PyArrow.
>> >>
>> >> ("PyCObject" is a Python 2-specific API that doesn't exist anymore in
>> >> Python 3)
>> >>
>> >> Regards
>> >>
>> >> Antoine.
>> >>
>> >>
>> >> Le 14/05/2018 à 18:34, Quang Vu a écrit :
>> >>> Hi Arrow dev,
>> >>>
>> >>> I am having trouble with installing and setting my development
>> >> environment
>> >>> for Arrow. I wonder if anyone is familiar with the issue. My system
>> info:
>> >>> - MacOS 10.11.6 (El Capitan)
>> >>> - conda 4.5.1
>> >>> - python 3.6.5
>> >>> - arrow's current commit: 4b8511
>> >>>
>> >>> Installing Arrow C++ libraries and Pacquet are both successful. But
>> >>> importing `pyarrow` fail:
>> >>>
>> >>> $ python -c 'import pyarrow'
>> >>>
>> >>> Traceback (most recent call last):
>> >>> File "<string>", line 1, in <module>
>> >>> File "/Users/myuser/code/arrow/python/pyarrow/__init__.py", line 47,
>> in
>> >>> <module>
>> >>> from pyarrow.lib import cpu_count, set_cpu_count
>> >>> ImportError: dlopen(/Users/myuser/code/arrow/python/pyarrow/
>> >>> lib.cpython-36m-darwin.so, 2): Symbol not found: _PyCObject_Type
>> >>> Referenced from:
>> >>> /Users/myuser/miniconda3/envs/pyarrow-test/lib/libarrow_python.10.dylib
>> >>> Expected in: flat namespace
>> >>> in
>> >> /Users/myuser/miniconda3/envs/pyarrow-test/lib/libarrow_python.10.dylib
>> >>>
>> >>> If anyone have suggestion on what the problem is about, please let me
>> >> know.
>> >>> Thanks!
>> >>>
>> >>
>> >
>>
